consider a variation of the "WAV to MIDI" conversion problem. I know the complexity of such a problem, and I know that there is extensive literature on a more general subject of music information retrieval (MIR).
But suppose we already have both a WAV and a MIDI representation of a piece of music, so we really don’t need to detect resins inside the WAV signal from scatch ... we just have to match the detected resins (using a suitable algorithm) with NoteOn events, contained in the MIDI view. I definitely believe that we should use the information contained in the MIDI file to give some hints to the pitch algorithm.
Such a matching tool can be very useful, for example, for “humanizing” MIDI: we could make the MIDI representation more expressive, using the information obtained from the WAV signal for hinting, duration, dynamics, etc.
Does anyone know if this problem has already been addressed in the literature? Any form of input or assistance would be greatly appreciated.
Thanks in advance.
source
share